Speaker: Cheng Tan (Northeastern University)
Time: 11:00 am, November 03 (Wednesday), 2021
Location: N/A
Online link: provided upon request or see the seminar email.
Abstract:
Neural networks (NNs) are beneficial to many applications and services, and we believe computer systems—such as OSes, databases, networked systems—are not an exception. However, applying NNs in these critical systems is challenging: people have to risk getting unexpected outcomes from NNs since NN behaviors are not well-defined.
In this talk, I will introduce our solution: certified neural network, a network that satisfy user-defined safety properties (called specifications). To build certified NNs, we introduce a system ouroboros which enables system developers to train networks that follow user-defined specifications. We do a case study on database learned indexes to demonstrate that training certified NNs is possible. Though many challenges remain, ouroboros enables us, for the first time, to apply NNs in critical systems with confidence.
Bio:
Cheng Tan is an assistant professor of Khoury College of Computer Sciences at Northeastern University. His research interests are in systems and security, with a focus on building verifiable outsourced services. His work has won SOSP’17 best paper award and Janet Fabri Prize for Outstanding Dissertation.